Within the sustainable development goal ‘climate action framework,’ forests play a vital role in enhancing climate resilience. The data obtained from these ecosystems are expected to significantly contribute to shaping regional forestry policies and advancing climate-related studies. Above-ground biomass (AGB) is a substantial parameter of the ecosystem cycle and is used to estimate forest carbon storage. Knowledge of the spatial distribution of AGB supports the management of forests and understanding of climate change’s effects. This study aims to analyze the relationship and to estimate biomass using machine learning methods between the AGB of a black pine (Pinus nigra) plantation forest obtained from ground measurements and the global mosaic L-band ALOS-PALSAR Synthetic Aperture Radar (SAR) image. The forest is distributed in western Inner Anatolia to the Black Sea, Marmara, and inner Aegean regions of Türkiye. Fieldwork was carried out in 44 plots distributed across different locations in Türkiye. The mean diameter, mean height, stand age, stand stock, stocking density, and basal area were determined for each sample. The aboveground tree masses of the sample trees were calculated. The relationship between ground-based AGB calculation and remote sensing ALOS data was obtained. This study estimated the AGB using linear regression, Random Forest, XGBoost, and Convolutional Neural Network (CNN) techniques. In addition to backscatter images of dual polarimetric (HH and HV) data, we created Normalized Difference Polarization Index (NDPI) and Co-polarized Ratio (CPR) images. Twelve scenarios were created using these four variables of ALOS data and their AGB estimation results were compared. . The highest AGB estimation performance obtained with the CNN model was achieved by combining HV polarization and CPR features, resulting in an R² value of 0.448 and an RMSE of 59.726 t/ha. The findings indicate that HV polarimetric data and integrating additional features further improved estimation capability. Overall, the findings demonstrate that integrating SAR imagery with CNN-based approaches provides an efficient, economically viable framework for multi-temporal forest biomass estimation. Such an approach has considerable potential to support large-scale forest carbon monitoring, sustainable forest management practices, and climate change mitigation policies at both national and global scales, including in Türkiye.